David, I think I chased this one down.  Here's a note from the archives...
give it a try.

Greg


I've recently run down a problem that can cause a hang when you try to
execute extremely large networks that involves a deadlock situation in the
socket communication between the exec and the UI.  To get around it I've
added an environment variable that allow you to specify a larger socket
buffer size (if the setsockopt call is available), named DX_SOCKET_BUFSIZE.
The value assigned to it is passed as the parameter to setsockopt for
SOL_SNDBUF and SOL_RCVBUF.  I think the default is 32767, and the range is
clamped to the max allowed by the system.   The value used is clamped to
the max allowed by the system; see your sysadmin if you need up this.  I'm
using 262144.

Greg

David Thompson <[EMAIL PROTECTED]>@opendx.watson.ibm.com on 03/01/2002
03:57:26 PM

Please respond to [email protected]

Sent by:    [EMAIL PROTECTED]


To:    [email protected]
cc:
Subject:    [opendx-dev] Packet problem



I've noticed going through the samples, one right after another quite
quickly--that once in a while, I get into a race condition in
communication between dxui and the exec. I just attached to one of
this hung execs and interrupted, and it said it was stuck at
sfile.c:191, which is a write. I will try and track this down
further, but has anybody else seen this behavior? This is going to be
one of those tough ones to nail.

David
--
.............................................................................

David L. Thompson                   Visualization and Imagery Solutions,
Inc.
mailto:[EMAIL PROTECTED]    5515 Skyway Drive, Missoula, MT 59804
                                     Phone : (406)257-8530


Reply via email to